OpenAI publishes updated Model Spec with five core values

OpenAI released an updated Model Spec, its public document on how models should behave. It outlines five core values meant to guide more capable AI toward sound judgment, human agency, and faithfulness to human intent.

Why it matters: The spec is the reference point for how OpenAI trains and evaluates behavior, so changes signal where it is shifting emphasis as models act more autonomously. The focus on judgment and human agency fits growing concern about agents acting beyond user intent, as in recent red-team and influence findings.